Remote SensingWetlands Delineation
Image Classification, Remote Sensing techniques, Lidar
The client had a need to update the existing wetlands data and extend it to the areas that were not covered during previous surveys. The objective was to run a desktop analysis by utilizing the latest available lidar elevation and aerial imagery datasets. Tessellations role on this wetlands boundary delineation project was focused on testing and applying various image processing techniques performed within ArcGIS and ERDAS imagine software.
Image processing and enhancement
Image classification and change detection analysis
Training samples and editing
Lidar data processing
Data documentation and reporting
- A hybrid approach as referred in publication from Lane et al. was utilized for this process. In this approach we used a composite of World View-3 imagery bands (R, G, B, NIR) along with texture and NDVI for improving the image classification output.
- An unsupervised classification method was run on the composite image and the output was reclassified into wetlands/non-wetlands based on visual interpretation.
- The output was further refined by cleaning the speckles using several iterations of image smoothing low pass filters.
- Lidar derived elevation data was also used to refine the boundary.
- The raster/image classification output was finally converted to a vector format (polygon feature class). Final cleanup was done by manually editing the feature class by visual interpretation of the imagery and by utilizing lidar elevation as the reference layers.